Science fiction movies of artificial intelligence (AI) are abound with plots of benign robots and computers that suddenly gain sentience and emotions, leading them to destroy all of humanity.
The seemingly helpful HAL in “2001: A Space Odyssey” is suffocating astronauts in their sleep by the end of the movie.
The latest in a long history of murderous robots, “Ex Machina’s” Ava kills one man and leaves another for dead in her quest to escape.
But these stories of emotionally unstable robots breed baseless fears about what our AI future will really be like. That future robots will have human-like emotions is a huge misconception, said Yann LeCun, the director of Facebook’s Artificial Intelligence Research team.
LeCun explains to Tech Insider via email how robot emotions will actually work.
Myth #1: Advanced robots will have feelings.
The AI we have right now don’t look anything like Ava, they look more like the specialised robots of Wall-E. Ava is what’s called artificial general intelligence (AGI), an AI that exhibits human-level intelligence on as many different tasks as an average human. She can see, speak, listen, reason, and even manipulate.
The AI we have now are what’s called artificial narrow intelligence (ANI), or AI that are amazing at very narrow, specialised tasks, like trading stocks or solving geometry. Moving forward, researchers will likely develop more sophisticated ANI for many different tasks — at least in the near future.
“Most AIs will be specialised and have no emotions,” LeCun told Tech Insider. “Your car’s auto-pilot will just drive your car,” and it won’t be programmed to “feel” any certain way about that.
Myth #2: Robots will develop emotions spontaneously.
But even Wall-E got it wrong. He was an ANI — a garbage compactor — that developed a lot of emotions. He experiences awe, fear, and love — emotions that don’t make him more efficient at his task. In reality, AI will only have emotions if they’re programmed with them. It wouldn’t be a byproduct of creating super-intelligent programs, because it wouldn’t be of any use.
The one reason we might program emotions into robots would be to make them easier to work with — to make them less like unfeeling automatons and more like a favourite coworker.
“We can build into them altruism and other drives that will make them pleasant for humans to interact with them and be around them,” LeCun wrote.
LeCun points to the AI character Samantha in “Her” as “not entirely implausible” because she’s built specifically to feel emotions — namely love — to build a relationship with her human partner. Still, LeCun said an AGI that you see in “Her” is still decades away.
But that wouldn’t include negative (and, in science fiction, dangerous) emotions like anger and jealousy.
Myth #3:Robot emotions will be similar to human emotions.
If humans program robots with emotions it’s likely they will look nothing like ours, LeCun said. Their emotions will be very rudimentary in comparison to humans. AI’s emotions will more reflect their programmed goals, based on the “anticipation of rewards,” according to LeCun.
“Right now, the way we train machines is ‘supervised,’ a bit like when we show a picture book to a toddler and tell them the name of everything,” LeCun said.
In this case, the AI system takes the role of a student while the researcher is the teacher. And like a human student, the AI would have an built-in drive to succeed, and would work to anticipate improved success.
But LeCun assures us that even if AI have some form of emotions, there’s no reason to fear. LeCun said that this goal-reward behaviour will likely be the whole scope of an AI’s emotional depth. The most destructive emotions like greed and anger, will remain uniquely human.
“There is no reason for AIs to have self-preservation instincts, jealousy,” LeCun said. “AIs will not have these destructive ’emotions’ unless we build these emotions into them. I don’t see why we would want to do that.
Business Insider Emails & Alerts
Site highlights each day to your inbox.